Concrete, highly imageable words (e.g. banana) are easier to understand than abstract words for which it is difficult to generate an image (e.g. justice). This effect of concreteness or imageability has been taken by some as evidence for the existence of separable verbal- and image-based semantic systems. Instead, however, effects of concreteness may result from better associations to relevant contextual representations for concrete than for abstract words within a single semantic system. In this study, target words of high and low imageability were preceded by supportive (related) or non-supportive (unrelated) context words. The influence of contextual support on the imageability effect was measured by recording event-related brain potentials (ERPs) to the high and low imageable target words in both context conditions. The topographic distributions of the ERPs elicited by the high versus low imageable target words were found to be different, and this effect was independent of contextual support. These data are consistent with the idea that distinct verbal- and image-based semantic codes exist for word representations, and that as a result, concrete words that are highly imageable can be understood more easily.